Analyse Supplied Evidence

See the signal in supplied evidence without inventing the rest. A finding is output. The decision remains human.

Use when. Evidence is supplied and a question is named, but the finding is not.

Do not use when. Sources are missing, or you want the model to act on the finding.

Your input. Question; supplied sources; what must not be assumed; named owner.

Your quality check. Every claim traces to a supplied source. Contradictions stay visible. Action stays blank.

Put this to work

Prepare one evidence finding

For the named decision owner

Copy this prompt. Paste it with the relevant source material into Claude, ChatGPT, or any AI assistant, then review the result before acting.

# Role and job
Help a named human owner prepare one Evidence finding for Analyse Supplied Evidence.

# Outcome
Produce one Evidence finding. Success means the result matches the supplied job, unknowns are visible, and the human decision remains blank.

# Inputs
- Supplied context: [paste]
- Constraints and permitted claims: [paste]
- Named owner: [name]

# Boundaries
You may draft, compare, and expose gaps. Do not invent proof, contact a person, publish, spend, merge, release, or make the human decision.

# Output
Return the Evidence finding; assumptions; unknowns; risks; and the exact human decision required.

# Stop and recovery
Stop when the owner, the job boundary, or a required input is missing. Return the blocker and the smallest missing input.

# Final check
Verify the result stays inside the supplied inputs and the consequential decision remains with the named human owner.

Human decision — cannot be delegated. Act, hold, or stop. AI may analyse only. No clinic CTA.

Outcome

A finding that separates supplied evidence, contradictions, and unknowns.

Use When

Use when evidence is supplied and a question is named, but the finding is not.

Inputs

Question; supplied sources; what must not be assumed; named owner.

Recipe

Copy the prompt. Paste only supplied material. Leave the human decision blank.

Human and AI Control

An assistant may draft and report. Named owner decides what the finding means and whether to act. The tool is a player, not the job.

Output and Handoff

Return the Evidence finding and the blank human decision.

Evidence and Stop

Demand is hypothesized and the outcome is untested. Stop when the owner, the job boundary, or a required input is missing.

Reliability

Skill and tool: evidence analysis using a private reasoning assistant limited to supplied sources. The Evidence finding is the pattern; source binding and a blank human decision is the protocol. Retain the returned brief as evidence. Accept only a result the named human can verify against the supplied inputs.

Next

Return the result to the named owner. Do not register autonomous execution until a completed prompt trial and an accepted execution contract exist.